Nowadays, portable computers equipped with wireless communication devices are becoming powerful enough to be used as ``clients on the road'', taking over tasks traditionally performed by desktop workstations. Having the luxury of large storage devices, they can cache significant amounts of data locally, and enable prolonged autonomous operation, i.e. the ability to work while disconnected from data servers. But, this also brings about the issue of data ``staleness'' and the need for frequent refresh of it. Examples include updating locally materialized views of database clients, and retrieving the latest news from information feeds. In this paper, we address the problem of propagating updates to a large number of mobile clients. We capitalize on the idea of adaptive hybrid (broadcast/unicast) data delivery to efficiently disseminate logged updates to mobile clients. First, we study the performance of log broadcasting using a multi-level air-cache structure. Then, we propose a set of techniques that dynamically modify the air-cache according to the size and the (dis)connection habits of the client population. Through extensive simulations, we demonstrate that the proposed system achieves scalable dissemination of updates, and show that, in all cases, hybrid delivery outperforms the best broadcast-only approach.

Back to the Spring 1998 dbchat index.